Volatility clearly affects profits. Returns per trade are smaller when volatility is low, and larger when it’s high. So is risk. But do trend systems perform better in periods of high or low volatility? The answer to that is important because it will tell you whether you should leverage down or up during those phases, or avoid them completely. The anecdotal evidence is that returns are better with high volatility, although maybe they are more consistent during low vol. That fact is that, personally, I’m not sure, but if one is better than the other, it will be important for trading.
Before going further, note that I am going to omit a great deal of detail and show some charts that will explain the process and the results. Those interested in using these results will need to run these same tests themselves anyway. This will help us get to the conclusions faster.
To keep it simple, we’ll look at the ETF SPY (SPDRs), from 1994 using a simple moving average system, with a calculation periods of 30-, 60-, and 120-days. Then we’ll do the same thing for Eurodollar interest rate futures. Trend systems are the most popular way of systematic trading, and 30-60-120 days covers a fair distribution of the periods, intended to capture the “macrotrend” domain.
There is only one rule in the trend system – we buy or sell when the trendline turns up or down. The size of the stock position is always equal to the investment ($10,000) divided by the price at the time of entry. For futures the investment is $25,000. We then have a modest amount of volatility parity.
We’ll look at both long and short sales because I believe that a system that profits best on both sides of the market is more robust and less risky. We can still choose to trade only the long side of stocks when we’re all done.
Applying the 60-day moving average to the 20 years of SPY data, we get the returns, shown as an NAV, in Chart 1. Stocks have had a volatile period since 1994, starting with the run-up to the tech bubble of 2000, followed by a bear market, another rally before the subprime crisis of 2008, and now a sustained 5-year bull market. These varying conditions are a good test for our purposes.
We’re going to be interested in the volatility of prices as well as the volatility of returns (from the NAVs). Both volatility calculations will be measured over 20 days, the same as most options, to avoid too many combinations.
Chart 1. SPY NAVs from a 60-Day moving average trading system.
Chart 2 shows price volatility and NAV volatility along with the prices of SPY.The points to recognize are that price and return vol (NAVvol) are very similar, and high vol occurs at key turning points in the SPY. During the 1990s, the bull market shows slowly rising volatility, but the rally from 2003 to 2007 does not, and the last five years shows erratic volatility. It will be a challenge to find consistency in these patterns; but the extreme high-vol periods look like low-hanging fruit.
Chart 2. Comparison of SPY price with volatility measured by price returns and NAV returns.
Chart 3. SPY performance with price and NAV filters.
Chart 4. SPY performance with low volatility price filter.
EURODOLLAR INTEREST RATE COMPARISON
Keeping in mind the results of the SPY tests, we’ll look at Eurodollar interest rate futures using typical back-adjusted data. That data makes it technically incorrect to use a percentage of prices, as with the SPY. In addition, the daily returns are not easily represented as percentages; therefore, we’ll substitute the daily profits/losses, which are not altered by the back-adjusting and calculate the volatility of performance. We’ll still use price returns, although yield returns would be better.
Chart 5 shows the familiar price of Eurodollar futures along with the NAVs resulting from simple 30-, 60-, and 120-day moving average systems. Those periods were chosen to be a fair sample of the “macrotrend.” Slower trends are expected to emphasize the consistency of the interest rate move over the past 20 to 30 years. In fact, due to that consistency, the fastest of the calculation periods netted the best returns.
The NAVs closely track the price patterns, flattening out recently as yields on Eurodollars hit all-time lows and price movement decreased to near zero. Chart 6 shows only the volatility, measured in the same way as we did earlier for the SPY, even though prices are back-adjusted, and measured from the performance NAVs. As with the SPY, both volatility measurements are very similar. In addition to the largest spikes occurring at the same time, both volatility lines decline to record lows beginning after 2008.
Chart 5. Moving average returns for the 30-, 60-, and 120-day periods.
Chart 6. Comparison of volatility for Eurodollars using the NAVs from the 120-day MA.
Using the same tests as we did with SPY, we find similar but much less noticeable improvements in the Eurodollar cumulative PL, seen in Chart 7. The high vol filter, based on daily changes in the PL, was best at 2.5% (see the left scale), and the filter based on a percentage of price change was best at 0.02. Although the percentage of PL is shown on the left scale and the percentage of price on the right scale, you can see that they are exactly at the same level, indicating that either measurement will achieve the same goal.
Chart 7. Results of high and low volatility filters on Eurodollars.
Although improvement in returns using the high-vol filter are barely noticeable, the ratio of returns to risk improve about 4% using the filter. That is not necessarily exciting, but it justifies the earlier result that high volatility does not have a positive effect on performance.
The low-vol tests gave similar, negative results, as those seen on SPY. There was no level where filtering the low volatility improved either the returns or the ratios. Chart 7 shows that the best low-vol filter blocked all trading in Eurodollars during the past two years.
Let’s say that the observations above can be generalized, that high vol is bad for performance and low vol is good. That conclusion seems reasonable to me given my market experience. Eliminating high-volatility trading may actually decrease net returns, but it reduces risk even more, and will be seen in the return/risk ratio. It is the reason why using other risk measures, such as VaR (value-at-risk) or VIX is justifiable. During the time that risk is extremely high, the returns are not usually enough to cover the increased risk. That should be a comforting conclusion because no one should want to be exposed to high risk.
IMPLICATIONS ON LEVERAGE
While reducing exposure to high risk is the obvious benefit, the lack of success finding a low volatility filter may be even more important. It means that returns under low volatility conditions are generally consistent. Now, combine that thought with the fact that daily returns during periods of low volatility are also low. Another way of looking at it is that we get steady, smaller returns during low-volatility intervals, which leads to profits but underperformance relative to our goals.
Professional money managers know that there is a concept called “volatility stabilization,” that changes the leverage in order to try to maintain a constant volatility of returns. In Managed Futures that target volatility is about 12%, but can vary from 8% to 16%, depending on the objective of the Manager.
In practice, if the volatility of returns is running at 8% and the target is 12%, then all positions must be increased by 50%. Of course, low volatility changes slowly, so that there is no one day where you would be buying 50% more. For futures these changes in leverage are easy, because there is always reserve capital with which to add or remove positions. For equities it is certainly more difficult. However, you can borrow at the current low rates and net an expected return far greater than the cost of capital, or you can “buy” leverage from your brokerage firm based on your risk profile and credit worthiness. Leverage can also be achieved using options.
The important point is that low volatility intervals have a better return ratio than high volatility periods, but low absolute returns. These can be leveraged up to take advantage of the performance profile.
© Copyright 2014, P.J. Kaufman. All right reserved.